I agree in principle, but Watson was a lot of different things, mostly unrelated to any “modern” AI tech. Edit: the Watson that played Jeopardy was smart (for

I agree in principle, but Watson was a lot of different things, mostly unrelated to any “modern” AI tech. Edit: the Watson that played Jeopardy was smart (for that time) natural language processing, that backed on to a large fact store, generated from natural language processing of a relatively small set of sources like Wikipedia. The Watson that was sold to companies was essentially a consulting service. I'm sure they had some re-usable components, but it's possible those components weren't much more than data pipelines, model training, etc. Most of the smarts were people doing a ton of data science specific to each problem. Worthwhile, but only good compared to companies who don't have a data science function. Conversely, "modern" AI is things like LLMs, which have their own problems of course, but are a huge paradigm shift compared to Watson.